Abstract: Due to selection difficulties for different bearing data feature, and low accuracy problems of single classifier method in the fault diagnosis of rolling bearing, this paper proposes a rolling bearing fault diagnosis algorithm with random forest based on Classification And Regression Tree（CART）. Random forest is an ensemble learning method which contains a variety of classifiers. The accuracy of rolling bearing fault diagnosis is improved by “integrated” thought of random forest. First, time domain statistical indicators are extracted from the vibration signals of rolling bearings and will be used as feature vectors. Then, the random forest algorithm is utilized for the fault diagnosis of rolling bearing. Compared with the traditional algorithm （SVM, kNN and ANN） and single CART, diagnostic results proposed in this paper indicate that random forest algorithm has high diagnostic accuracy by using the bearing data of SQI-MFS experimental platform.

Key words: rolling bearing, fault diagnosis, feature extraction, random forest

摘要： 针对不同轴承数据特征选择困难和单个分类器方法在滚动轴承故障诊断中精度较低的问题，提出了一种基于分类回归树（CART）的随机森林滚动轴承故障诊断算法。随机森林是包含了多种分类器的集成学习方法。通过随机森林的“集成”思想来提高滚动轴承故障诊断的精度。从滚动轴承的振动信号中提取时域统计指标，将其作为特征向量，利用随机森林（Random Forest）对滚动轴承故障进行诊断。利用SQI-MFS实验平台的轴承数据，与传统分类器（SVM、kNN和ANN）以及单个分类回归树的诊断结果相比，随机森林算法具有比较高的诊断精度。

关键词: 滚动轴承, 故障诊断, 特征提取, 随机森林